Index: lib/CodeGen/SelectionDAG/TargetLowering.cpp
===================================================================
--- lib/CodeGen/SelectionDAG/TargetLowering.cpp
+++ lib/CodeGen/SelectionDAG/TargetLowering.cpp
@@ -1426,12 +1426,36 @@
       if (SimplifyDemandedBits(Src, DemandedSrcBits, DemandedSrcElts,
                                KnownSrcBits, TLO, Depth + 1))
         return true;
+    } else if ((NumSrcEltBits % BitWidth) == 0 &&
+               TLO.DAG.getDataLayout().isLittleEndian()) {
+      unsigned Scale = NumSrcEltBits / BitWidth;
+      unsigned NumSrcElts = SrcVT.isVector() ? SrcVT.getVectorNumElements() : 1;
+      APInt DemandedSrcBits = APInt::getNullValue(NumSrcEltBits);
+      APInt DemandedSrcElts = APInt::getNullValue(NumSrcElts);
+      for (unsigned i = 0; i != NumElts; ++i)
+        if (DemandedElts[i]) {
+          unsigned Offset = (i % Scale) * BitWidth;
+          DemandedSrcBits.insertBits(DemandedBits, Offset);
+          DemandedSrcElts.setBit(i / Scale);
+        }
+
+      if (SrcVT.isVector()) {
+        APInt KnownSrcUndef, KnownSrcZero;
+        if (SimplifyDemandedVectorElts(Src, DemandedSrcElts, KnownSrcUndef,
+                                       KnownSrcZero, TLO, Depth + 1))
+          return true;
+      }
+
+      KnownBits KnownSrcBits;
+      if (SimplifyDemandedBits(Src, DemandedSrcBits, DemandedSrcElts,
+                               KnownSrcBits, TLO, Depth + 1))
+        return true;
     }
 
     // If this is a bitcast, let computeKnownBits handle it.  Only do this on a
     // recursive call where Known may be useful to the caller.
     if (Depth > 0) {
-      Known = TLO.DAG.computeKnownBits(Op, Depth);
+      Known = TLO.DAG.computeKnownBits(Op, DemandedElts, Depth);
       return false;
     }
     break;
Index: test/CodeGen/AMDGPU/store-weird-sizes.ll
===================================================================
--- test/CodeGen/AMDGPU/store-weird-sizes.ll
+++ test/CodeGen/AMDGPU/store-weird-sizes.ll
@@ -45,7 +45,8 @@
 ; HAWAII-NEXT:    v_mov_b32_e32 v2, s2
 ; HAWAII-NEXT:    ds_write_b16 v1, v2 offset:4
 ; HAWAII-NEXT:    s_waitcnt vmcnt(0)
-; HAWAII-NEXT:    v_and_b32_e32 v0, 0x7f, v0
+; HAWAII-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
+; HAWAII-NEXT:    v_bfe_u32 v0, v0, 16, 7
 ; HAWAII-NEXT:    ds_write_b8 v1, v0 offset:6
 ; HAWAII-NEXT:    ds_write_b32 v1, v3
 ; HAWAII-NEXT:    s_endpgm
@@ -67,7 +68,8 @@
 ; FIJI-NEXT:    v_mov_b32_e32 v2, s2
 ; FIJI-NEXT:    ds_write_b16 v1, v2 offset:4
 ; FIJI-NEXT:    s_waitcnt vmcnt(0)
-; FIJI-NEXT:    v_and_b32_e32 v0, 0x7f, v0
+; FIJI-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
+; FIJI-NEXT:    v_bfe_u32 v0, v0, 16, 7
 ; FIJI-NEXT:    ds_write_b8 v1, v0 offset:6
 ; FIJI-NEXT:    ds_write_b32 v1, v3
 ; FIJI-NEXT:    s_endpgm
Index: test/CodeGen/X86/bitcast-setcc-256.ll
===================================================================
--- test/CodeGen/X86/bitcast-setcc-256.ll
+++ test/CodeGen/X86/bitcast-setcc-256.ll
@@ -448,22 +448,6 @@
 define void @bitcast_4i64_store(i4* %p, <4 x i64> %a0) {
 ; SSE2-SSSE3-LABEL: bitcast_4i64_store:
 ; SSE2-SSSE3:       # %bb.0:
-; SSE2-SSSE3-NEXT:    movdqa {{.*#+}} xmm2 = [2147483648,2147483648]
-; SSE2-SSSE3-NEXT:    pxor %xmm2, %xmm1
-; SSE2-SSSE3-NEXT:    movdqa %xmm2, %xmm3
-; SSE2-SSSE3-NEXT:    pcmpeqd %xmm1, %xmm3
-; SSE2-SSSE3-NEXT:    movdqa %xmm2, %xmm4
-; SSE2-SSSE3-NEXT:    pcmpgtd %xmm1, %xmm4
-; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[0,0,2,2]
-; SSE2-SSSE3-NEXT:    pand %xmm3, %xmm1
-; SSE2-SSSE3-NEXT:    por %xmm4, %xmm1
-; SSE2-SSSE3-NEXT:    pxor %xmm2, %xmm0
-; SSE2-SSSE3-NEXT:    movdqa %xmm2, %xmm3
-; SSE2-SSSE3-NEXT:    pcmpeqd %xmm0, %xmm3
-; SSE2-SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
-; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[0,0,2,2]
-; SSE2-SSSE3-NEXT:    pand %xmm3, %xmm0
-; SSE2-SSSE3-NEXT:    por %xmm2, %xmm0
 ; SSE2-SSSE3-NEXT:    packssdw %xmm1, %xmm0
 ; SSE2-SSSE3-NEXT:    movmskps %xmm0, %eax
 ; SSE2-SSSE3-NEXT:    movb %al, (%rdi)
Index: test/CodeGen/X86/dagcombine-cse.ll
===================================================================
--- test/CodeGen/X86/dagcombine-cse.ll
+++ test/CodeGen/X86/dagcombine-cse.ll
@@ -14,18 +14,11 @@
 ;
 ; X64-LABEL: t:
 ; X64:       ## %bb.0: ## %entry
-; X64-NEXT:    ## kill: def $edx killed $edx def $rdx
-; X64-NEXT:    ## kill: def $esi killed $esi def $rsi
 ; X64-NEXT:    imull %ecx, %esi
-; X64-NEXT:    leal (%rsi,%rdx), %eax
-; X64-NEXT:    cltq
+; X64-NEXT:    addl %edx, %esi
+; X64-NEXT:    movslq %esi, %rax
 ; X64-NEXT:    movl (%rdi,%rax), %eax
-; X64-NEXT:    leal 4(%rsi,%rdx), %ecx
-; X64-NEXT:    movslq %ecx, %rcx
-; X64-NEXT:    movzwl (%rdi,%rcx), %ecx
-; X64-NEXT:    shlq $32, %rcx
-; X64-NEXT:    orq %rax, %rcx
-; X64-NEXT:    movq %rcx, %xmm0
+; X64-NEXT:    movq %rax, %xmm0
 ; X64-NEXT:    movd %xmm0, %eax
 ; X64-NEXT:    retq
 entry:
Index: test/CodeGen/X86/masked_store.ll
===================================================================
--- test/CodeGen/X86/masked_store.ll
+++ test/CodeGen/X86/masked_store.ll
@@ -36,25 +36,21 @@
 define void @store_v2f64_v2i64(<2 x i64> %trigger, <2 x double>* %addr, <2 x double> %val) {
 ; SSE2-LABEL: store_v2f64_v2i64:
 ; SSE2:       ## %bb.0:
-; SSE2-NEXT:    movdqa {{.*#+}} xmm3 = [2147483648,2147483648]
-; SSE2-NEXT:    pxor %xmm3, %xmm0
-; SSE2-NEXT:    movdqa %xmm3, %xmm2
-; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
-; SSE2-NEXT:    pcmpeqd %xmm3, %xmm0
-; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
-; SSE2-NEXT:    movdqa %xmm0, %xmm4
-; SSE2-NEXT:    pand %xmm2, %xmm4
-; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[1,1,3,3]
-; SSE2-NEXT:    por %xmm3, %xmm4
-; SSE2-NEXT:    movd %xmm4, %eax
+; SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [2147483648,2147483648]
+; SSE2-NEXT:    pxor %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm2, %xmm3
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm3
+; SSE2-NEXT:    pcmpeqd %xmm2, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[1,1,3,3]
+; SSE2-NEXT:    por %xmm2, %xmm0
+; SSE2-NEXT:    movd %xmm0, %eax
 ; SSE2-NEXT:    testb $1, %al
 ; SSE2-NEXT:    je LBB1_2
 ; SSE2-NEXT:  ## %bb.1: ## %cond.store
 ; SSE2-NEXT:    movlpd %xmm1, (%rdi)
 ; SSE2-NEXT:  LBB1_2: ## %else
-; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,0,2,2]
-; SSE2-NEXT:    pand %xmm2, %xmm0
-; SSE2-NEXT:    por %xmm3, %xmm0
 ; SSE2-NEXT:    pextrw $4, %xmm0, %eax
 ; SSE2-NEXT:    testb $1, %al
 ; SSE2-NEXT:    je LBB1_4
@@ -117,20 +113,16 @@
 ; SSE2-NEXT:    movdqa %xmm4, %xmm5
 ; SSE2-NEXT:    pcmpgtd %xmm0, %xmm5
 ; SSE2-NEXT:    pcmpeqd %xmm4, %xmm0
-; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
-; SSE2-NEXT:    movdqa %xmm0, %xmm7
-; SSE2-NEXT:    pand %xmm5, %xmm7
-; SSE2-NEXT:    pshufd {{.*#+}} xmm6 = xmm5[1,1,3,3]
-; SSE2-NEXT:    por %xmm6, %xmm7
-; SSE2-NEXT:    movd %xmm7, %eax
+; SSE2-NEXT:    pshufd {{.*#+}} xmm6 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm5, %xmm6
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[1,1,3,3]
+; SSE2-NEXT:    por %xmm6, %xmm0
+; SSE2-NEXT:    movd %xmm0, %eax
 ; SSE2-NEXT:    testb $1, %al
 ; SSE2-NEXT:    je LBB2_2
 ; SSE2-NEXT:  ## %bb.1: ## %cond.store
 ; SSE2-NEXT:    movlpd %xmm2, (%rdi)
 ; SSE2-NEXT:  LBB2_2: ## %else
-; SSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,0,2,2]
-; SSE2-NEXT:    pand %xmm5, %xmm0
-; SSE2-NEXT:    por %xmm6, %xmm0
 ; SSE2-NEXT:    pextrw $4, %xmm0, %eax
 ; SSE2-NEXT:    testb $1, %al
 ; SSE2-NEXT:    je LBB2_4
@@ -140,10 +132,9 @@
 ; SSE2-NEXT:    pxor %xmm4, %xmm1
 ; SSE2-NEXT:    movdqa %xmm4, %xmm0
 ; SSE2-NEXT:    pcmpgtd %xmm1, %xmm0
-; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[0,0,2,2]
 ; SSE2-NEXT:    pcmpeqd %xmm4, %xmm1
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; SSE2-NEXT:    pand %xmm2, %xmm1
+; SSE2-NEXT:    pand %xmm0, %xmm1
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
 ; SSE2-NEXT:    por %xmm1, %xmm0
 ; SSE2-NEXT:    pextrw $0, %xmm0, %eax
@@ -863,25 +854,21 @@
 define void @store_v2i64_v2i64(<2 x i64> %trigger, <2 x i64>* %addr, <2 x i64> %val) {
 ; SSE2-LABEL: store_v2i64_v2i64:
 ; SSE2:       ## %bb.0:
-; SSE2-NEXT:    movdqa {{.*#+}} xmm3 = [2147483648,2147483648]
-; SSE2-NEXT:    pxor %xmm3, %xmm0
-; SSE2-NEXT:    movdqa %xmm3, %xmm2
-; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
-; SSE2-NEXT:    pcmpeqd %xmm3, %xmm0
-; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
-; SSE2-NEXT:    movdqa %xmm0, %xmm4
-; SSE2-NEXT:    pand %xmm2, %xmm4
-; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[1,1,3,3]
-; SSE2-NEXT:    por %xmm3, %xmm4
-; SSE2-NEXT:    movd %xmm4, %eax
+; SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [2147483648,2147483648]
+; SSE2-NEXT:    pxor %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm2, %xmm3
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm3
+; SSE2-NEXT:    pcmpeqd %xmm2, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[1,1,3,3]
+; SSE2-NEXT:    por %xmm2, %xmm0
+; SSE2-NEXT:    movd %xmm0, %eax
 ; SSE2-NEXT:    testb $1, %al
 ; SSE2-NEXT:    je LBB7_2
 ; SSE2-NEXT:  ## %bb.1: ## %cond.store
 ; SSE2-NEXT:    movq %xmm1, (%rdi)
 ; SSE2-NEXT:  LBB7_2: ## %else
-; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,0,2,2]
-; SSE2-NEXT:    pand %xmm2, %xmm0
-; SSE2-NEXT:    por %xmm3, %xmm0
 ; SSE2-NEXT:    pextrw $4, %xmm0, %eax
 ; SSE2-NEXT:    testb $1, %al
 ; SSE2-NEXT:    je LBB7_4
@@ -950,20 +937,16 @@
 ; SSE2-NEXT:    movdqa %xmm4, %xmm5
 ; SSE2-NEXT:    pcmpgtd %xmm0, %xmm5
 ; SSE2-NEXT:    pcmpeqd %xmm4, %xmm0
-; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
-; SSE2-NEXT:    movdqa %xmm0, %xmm7
-; SSE2-NEXT:    pand %xmm5, %xmm7
-; SSE2-NEXT:    pshufd {{.*#+}} xmm6 = xmm5[1,1,3,3]
-; SSE2-NEXT:    por %xmm6, %xmm7
-; SSE2-NEXT:    movd %xmm7, %eax
+; SSE2-NEXT:    pshufd {{.*#+}} xmm6 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm5, %xmm6
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[1,1,3,3]
+; SSE2-NEXT:    por %xmm6, %xmm0
+; SSE2-NEXT:    movd %xmm0, %eax
 ; SSE2-NEXT:    testb $1, %al
 ; SSE2-NEXT:    je LBB8_2
 ; SSE2-NEXT:  ## %bb.1: ## %cond.store
 ; SSE2-NEXT:    movq %xmm2, (%rdi)
 ; SSE2-NEXT:  LBB8_2: ## %else
-; SSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,0,2,2]
-; SSE2-NEXT:    pand %xmm5, %xmm0
-; SSE2-NEXT:    por %xmm6, %xmm0
 ; SSE2-NEXT:    pextrw $4, %xmm0, %eax
 ; SSE2-NEXT:    testb $1, %al
 ; SSE2-NEXT:    je LBB8_4
@@ -974,10 +957,9 @@
 ; SSE2-NEXT:    pxor %xmm4, %xmm1
 ; SSE2-NEXT:    movdqa %xmm4, %xmm0
 ; SSE2-NEXT:    pcmpgtd %xmm1, %xmm0
-; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[0,0,2,2]
 ; SSE2-NEXT:    pcmpeqd %xmm4, %xmm1
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; SSE2-NEXT:    pand %xmm2, %xmm1
+; SSE2-NEXT:    pand %xmm0, %xmm1
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
 ; SSE2-NEXT:    por %xmm1, %xmm0
 ; SSE2-NEXT:    pextrw $0, %xmm0, %eax
Index: test/CodeGen/X86/movmsk-cmp.ll
===================================================================
--- test/CodeGen/X86/movmsk-cmp.ll
+++ test/CodeGen/X86/movmsk-cmp.ll
@@ -929,22 +929,6 @@
 define i1 @allones_v4i64_sign(<4 x i64> %arg) {
 ; SSE2-LABEL: allones_v4i64_sign:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [2147483648,2147483648]
-; SSE2-NEXT:    pxor %xmm2, %xmm1
-; SSE2-NEXT:    movdqa %xmm2, %xmm3
-; SSE2-NEXT:    pcmpeqd %xmm1, %xmm3
-; SSE2-NEXT:    movdqa %xmm2, %xmm4
-; SSE2-NEXT:    pcmpgtd %xmm1, %xmm4
-; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[0,0,2,2]
-; SSE2-NEXT:    pand %xmm3, %xmm1
-; SSE2-NEXT:    por %xmm4, %xmm1
-; SSE2-NEXT:    pxor %xmm2, %xmm0
-; SSE2-NEXT:    movdqa %xmm2, %xmm3
-; SSE2-NEXT:    pcmpeqd %xmm0, %xmm3
-; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
-; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[0,0,2,2]
-; SSE2-NEXT:    pand %xmm3, %xmm0
-; SSE2-NEXT:    por %xmm2, %xmm0
 ; SSE2-NEXT:    packssdw %xmm1, %xmm0
 ; SSE2-NEXT:    movmskps %xmm0, %eax
 ; SSE2-NEXT:    cmpb $15, %al
@@ -989,22 +973,6 @@
 define i1 @allzeros_v4i64_sign(<4 x i64> %arg) {
 ; SSE2-LABEL: allzeros_v4i64_sign:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [2147483648,2147483648]
-; SSE2-NEXT:    pxor %xmm2, %xmm1
-; SSE2-NEXT:    movdqa %xmm2, %xmm3
-; SSE2-NEXT:    pcmpeqd %xmm1, %xmm3
-; SSE2-NEXT:    movdqa %xmm2, %xmm4
-; SSE2-NEXT:    pcmpgtd %xmm1, %xmm4
-; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[0,0,2,2]
-; SSE2-NEXT:    pand %xmm3, %xmm1
-; SSE2-NEXT:    por %xmm4, %xmm1
-; SSE2-NEXT:    pxor %xmm2, %xmm0
-; SSE2-NEXT:    movdqa %xmm2, %xmm3
-; SSE2-NEXT:    pcmpeqd %xmm0, %xmm3
-; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
-; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[0,0,2,2]
-; SSE2-NEXT:    pand %xmm3, %xmm0
-; SSE2-NEXT:    por %xmm2, %xmm0
 ; SSE2-NEXT:    packssdw %xmm1, %xmm0
 ; SSE2-NEXT:    movmskps %xmm0, %eax
 ; SSE2-NEXT:    testb %al, %al
@@ -4522,22 +4490,6 @@
 define i32 @movmskpd256(<4 x double> %x) {
 ; SSE2-LABEL: movmskpd256:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [2147483648,2147483648]
-; SSE2-NEXT:    pxor %xmm2, %xmm1
-; SSE2-NEXT:    movdqa %xmm2, %xmm3
-; SSE2-NEXT:    pcmpeqd %xmm1, %xmm3
-; SSE2-NEXT:    movdqa %xmm2, %xmm4
-; SSE2-NEXT:    pcmpgtd %xmm1, %xmm4
-; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[0,0,2,2]
-; SSE2-NEXT:    pand %xmm3, %xmm1
-; SSE2-NEXT:    por %xmm4, %xmm1
-; SSE2-NEXT:    pxor %xmm2, %xmm0
-; SSE2-NEXT:    movdqa %xmm2, %xmm3
-; SSE2-NEXT:    pcmpeqd %xmm0, %xmm3
-; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
-; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[0,0,2,2]
-; SSE2-NEXT:    pand %xmm3, %xmm0
-; SSE2-NEXT:    por %xmm2, %xmm0
 ; SSE2-NEXT:    packssdw %xmm1, %xmm0
 ; SSE2-NEXT:    movmskps %xmm0, %eax
 ; SSE2-NEXT:    retq